The Gap Between "Seems Hard to Break" and Provably Hard to Break
Before provable security, cryptographic schemes were evaluated by how long cryptanalysts failed to break them. Modern cryptography does better: it defines security precisely as a game, then proves that breaking the scheme is at least as hard as solving a recognized mathematical problem. This does not eliminate assumptions — it makes them explicit and minimal.
The Security Game Paradigm
Security is defined as a game between an adversary A and a challenger C. The challenger holds a secret (a key, a random choice). The adversary queries the challenger and attempts a task — distinguish two ciphertexts, forge a signature, recover a plaintext
IND-CPA: Semantic Security
Indistinguishability under Chosen Plaintext Attack (IND-CPA): the adversary submits two messages of equal length, the challenger encrypts one chosen at random, and the adversary must identify which was encrypted. A semantically secure cipher gives the adv
IND-CCA2: The Gold Standard for Encryption
Indistinguishability under Adaptive Chosen Ciphertext Attack (IND-CCA2) gives the adversary a decryption oracle: it can request decryption of any ciphertext except the challenge ciphertext. Schemes secure under IND-CCA2 are robust against active attackers
Reduction Proofs
To prove scheme S is IND-CPA secure, construct a reduction R: a polynomial-time algorithm that uses any adversary A breaking S as a subroutine to solve the underlying hard problem. If A breaks S with non-negligible probability, R solves the hard problem w
The Random Oracle Model and the Standard Model
Many security proofs idealize hash functions as truly random functions available to all parties — the Random Oracle Model (ROM). ROM proofs are cleaner and allow tighter security bounds. They are criticized because real hash functions are not random oracl
Go Deeper: Security Reductions and Computability Theory
Security reductions work in the asymptotic setting — security grows as a function of the security parameter. Making that precise requires the formal language of polynomial-time computation and computability: what functions can be computed and what cannot.
